Climate risks and U.S. stock-market tail risks : a forecasting experiment using over a century of data

Abstract

We examine the predictive value of the uncertainty associated with growth in temperature for stock-market tail risk in the United States using monthly data that cover the sample period from 1895:02 to 2021:08. To this end, we measure stock-market tail risk by means of the popular Conditional Autoregressive Value at Risk (CAViaR) model. Our results show that accounting for the predictive value of the uncertainty associated with growth in temperature, as measured either by means of standard generalized autoregressive conditional heteroskedasticity (GARCH) models or a stochastic-volatility (SV) model, mainly is beneficial for a forecaster who suffers a sufficiently higher loss from an underestimation of tail risk than from a comparable overestimation.